N E W S A N D V I E W S

512 VOLUME 4 NUMBER 6 JUNE 2003 NATURE IMMUNOLOGY

Kruppel-like transcription factors1, such asthe Ikaros family of proteins and the lungKruppel-like transcription factor, are essen-tial for lymphoid development2 and T cellhomeostasis3. Two Kruppel-like transcrip-tion factors, Bcl11a and Bcl11b, have nowbeen shown to be required in lymphoiddevelopment based on the phenotype thatresults from knockout of the correspondinggenes, as described by Liu et al.4 andWakabayashi et al.5, respectively, in this issueof Nature Immunology. The block in lym-

phoid development is marked in these miceand introduces two new key players and aquest to find out how they work.

Bcl11a and Bcl11b are similar in sequenceand physical properties. Both proteins canact as transcriptional repressors in transienttransfection assays and both bind the sameconsensus sequence and repress transcrip-tion of a reporter gene6. Both Bcl11a andBcl11b also bind to a type of orphan nuclearreceptor (COUP-TF1) that itself repressestranscription. But transcriptional repressionmay not be their key function, because otherC2H2 zinc finger proteins, like Ikaros andYY1, can also function as transcriptionalactivators in some promoters. Mice in whicheither Bcl11a or Bcl11b has been knocked out

die shortly after birth of unknown causes,and both types of knockout mice have defectsin early lymphoid development4,5. Despitethe structural similarity of Bcl11a andBcl11b, however, the lymphoid defects thatresult from knocking them out are distinct(Fig. 1). Whereas Bcl11a is required for B celldevelopment, Bcl11b is required for αβ celldevelopment.

Bcl11a was first discovered in the mouse asa site of retroviral integration and termedEvi9. It was activated by viral integration, itinduced myeloid leukemias and it wasshown to transform NIH 3T3 cells7,8. Themouse gene was also independently identi-fied as a protein that interacted with thechicken ovalbumin upstream promoter

The author is with the National Cancer Institute,

National Institutes of Health, Bethesda, Maryland

20892, USA. e-mail: durums@mail.ncifcrf.gov

Bcl11: sibling rivalry in lymphoid developmentScott K Durum

Lymphocyte development is a complex process that involves the coordinated action of many transcription factors. Thepuzzle of B cell and T cell development gains an additional clue with the discovery of two critical factors.

data of Willert et al.8 indicating that Wnt3exposure increased the frequency of cloneswith HSC activity as compared to the fre-quency of HSCs in the starting population.It might thus be speculated that reinstate-ment of sufficient levels of activated β-catenin conferred HSC status again on latertypes of hematopoietic cells that were stillpluripotent but already molecularly des-tined to complete the changes that establishlineage restriction within a few cell cycles.An intermediate population of pluripotenthematopoietic cells has been well recog-nized for many years, and these cells areknown to differ from HSCs in terms of theirphenotype and regulation as well as theirengrafting durability.

A model of HSC regulation might beenvisaged in which the HSC state includesthe acquisition of competence to differenti-ate into all of the blood cell lineages as wellas a mechanism to block the activation ofevents that initiate these differentiationprocesses (Fig. 1). It is interesting to notethat, during embryogenesis, the firsthematopoietic cells to appear arise directlyfrom mesodermal precursors without pass-ing through a state that would meet thefunctional definition of a transplantableHSC with self-renewal activity. In fact,such cells develop relatively late andbecome detectable only after pluripotent

hematopoietic cells can already be identi-fied. These findings are consistent with theconcept that HSC development representsa molecularly distinct process requiring theactivation of a mechanism for blocking theexpression of a pre-established differentia-tion potential during successive cellcycles—a property that would then bedetected functionally as self-renewal.

A number of transcription factorsthought to be involved in the initial estab-lishment of hematopoietic competence andin the final control of lineage programminghave been defined. Transcription factorsthat might have comparable roles in regu-lating HSC self-renewal are newer to ourthinking. As recognized in the study byReya et al., however, enhanced expressionof Hoxb4, previously shown to characterizeHSC-enriched populations, and the pro-found ability of forced overexpression ofHoxb4 to stimulate HSC amplification invitro and in vivo11–13 strongly point to thisgene as a primary candidate. It was there-fore provocative to observe that theincreased HSC activity resulting fromforced overexpression of activated β-catenin in HSC-enriched populations wasparalleled by a specific enhancement inHoxb4 expression.

Identification of a new method for isolat-ing bioactive Wnt proteins and inducing

downstream activation of the Wnt–β-catenin pathway in HSCs raises many excit-ing possibilities for future exploitation inthe genetic manipulation and transplanta-tion of patient HSC populations. At thesame time, the association of Wnt1 overex-pression with human leukemia suggestscaution in how these are pursued and high-lights the need for a better understanding ofthe fine line between normal and neoplasticstem cell behavior.

1. Ivanova, N.B. et al. Science 298, 601–604 (2002).2. Ramalho-Santos, M., Yoon, S., Matsuzaki, Y.,

Mulligan, R.C. & Melton, D.A. Science 298,597–600 (2002).

3. Miller, C.L. & Eaves, C.J. Proc. Natl. Acad. Sci. USA94, 13648–13653 (1997).

4. Yonemura, Y., Ku, H., Hirayama, F., Souza, L.M. &Ogawa, M. Proc. Natl. Acad. Sci. USA 93,4040–4044 (1996).

5. Ema, H., Takano, H., Sudo, K. & Nakauchi, H. J.Exp. Med. 192, 1281–1288 (2000).

6. Audet, J., Miller, C.L., Eaves, C.J. & Piret, J.M.Biotechnol. Bioeng. 80, 393–404 (2002).

7. Reya, T. et al. Nature advance online publication,27 April 2003 (doi:10.1038/nature01593).

8. Willert, K. et al. Nature advance online publication,27 April 2003 (doi:10.1038/nature01611).

9. Szilvassy, S.J., Ragland, P.L., Miller, C.L. & Eaves,C.J. Exp. Hematol. 31, 331–338 (2003).

10. Iscove, N. et al. Exp. Hematol. 30 (Suppl 1), 38(2002).

11. Sauvageau, G. et al. Proc. Natl. Acad. Sci. USA 91,12223–12227 (1994).

12. Sauvageau, G. et al. Genes Dev. 9, 1753–1765(1995).

13. Antonchuk, J., Sauvageau, G. & Humphries, R.K.Cell 109, 39-45 (2002).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reim

mu

no

log

y

transcription factor9. The human ortholog,BCL11A, was discovered10 as a rare translo-cation involving the IGH locus,t(2;14)(p13;q32.3), in aggressive B cellchronic lymphocytic leukemias. The sameregion (2p13) is commonly overexpressed inB cell lymphomas and Hodgkin disease.Thus, BCL11A seems to be a proto-oncogenein lymphocytes in humans and myeloid cellsin mice.

The Bcl11a protein is found in thenucleus, and there is evidence it can act as atranscriptional repressor6. A consensusDNA sequence to which it binds was identi-fied, but the physiological target genes ofBcl11a are unknown. It can interact withanother Kruppel-type zinc finger protein,encoded by BCL6, which is a proto-onco-gene in B cells and also can be a sequence-specific transcriptional repressor11. It alsobinds with members of the steroid-thyroidhormone receptor superfamily9.

As shown by Liu et al., Bcl11a is clearlyrequired for B cell development4, consistentwith the observation in humans that overex-pression seems to promote B cell oncogene-sis. The profound block in B celldevelopment in the absence of Bcl11aoccurred before the earliest stages of expres-sion of Ebf, Pax5, Il7Ra and Cd19. Althoughretroviral activation of Bcl11a inducedmyeloid tumors in mice, there was no defectin myeloid development in knockout mice.

Thymic development was also perturbed inBcl11a knockouts, but not to the same degreeas B cell development. The thymic defect maybe due in part to a reduction in the commonlymphoid progenitors, as the thymus wasone-quarter to one-third the size in some,but not all, of the mice and transfer of prog-enitors poorly repopulated the recipient thy-mus. The number of γδthymocytes, however,was not lower, so there may also be a specificrole for Bcl11a in αβ thymocytes, especiallyin the CD4 lineage, which was reduced innumbers.

As suggested by the authors, Bcl11a couldregulate Notch1 signaling. Notch1 has beenshown to inhibit B cell development and pro-mote T cell development. If Bcl11a inhibitsthe Notch1 pathway, then knockout of Bcl11amight release the brake on Notch1, whichwould in turn block B cell development.There is no direct evidence for this model,but large amounts of Notch1 transcripts werenoted in the thymic lymphomas that devel-oped from host cells in recipients of Bcl11a–/–

stem cells. The effect of Bcl11a could also beon ligands for Notch1. A prediction of thismodel is that Notch1 deficiency should res-cue B cell development in Bcl11a–/– mice.

A notable finding is the evidence for a‘non-autonomous tumor suppressor func-tion’ for Bcl11a. The authors suggest thatBcl11a–/– hematopoietic cells can inducenormal thymocytes to transform. Support

for this hypothesis came from their experi-mental observation that Bcl11a–/– bone mar-row induces thymic lymphomas of hostorigin three months after transfer into alethally irradiated normal mouse. If otherknown tumor suppressor genes regulatetransformation by controlling cell growthand death from inside the cell, how couldBcl11a suppress transformation in anothercell? There is some recent evidence of tumorsuppression by lymphocytes and interferon-γ, so perhaps Bcl11a somehow promotes thisprocess. Or Bcl11a could induce a memberof the TGF-β or TNF families of ligands,which can regulate lymphocyte growth ordeath from the outside, as shown by thetransformation of neighboring cells by α-inhibin knockout. Another possibility is thatBcl11a inhibits production of survival orproliferation factors for lymphocytes.Examples of such factors are IL-6 and IL-7,which, if produced in excess (for example, astransgenes or by viruses), can induce anaccumulation of lymphocytes and con-tribute to their transformation. Notch lig-ands also would work non-autonomouslyand can protect thymocytes from apoptosis.Thus, Bcl11a might normally induce synthe-sis of a survival or growth inhibitor, it mightpromote immune surveillance or it mightinhibit the synthesis of a survival or growthstimulant.

The transformed thymocytes in Bcl11a–/–

recipients are at the immature double-posi-tive stage, a point at which knockout andtransgenic mice with high levels of DNAdamage or growth commonly developtumors. Why thymocytes at this stage are sovulnerable is unclear; it is probably notbecause VDJ recombination errors activateoncogenes but rather because thymocytesproliferate dangerously quickly. One con-cern about the non-autonomous tumor sup-pression interpretation is that theexperimental result could possibly beexplained by endogenous retroviruses.Bcl11a might repress expression of anendogenous retrovirus. For example, irradi-ation can awaken leukemia viruses that theninfect neighboring cells. The authors notethat Bcl11a–/– mice on a mixed strain back-ground have tumor-inducing capacitywhereas those on the C57Bl/6 backgrounddo not, indicating that there is a cofactor,which could be an endogenous virus.

The paralog of BCL11A, BCL11B, was dis-covered in human databases and is mappedto a different chromosome (14q32.1; ref.10). The mouse ortholog of BCL11B5 wasdiscovered independently by analyzing aregion of loss of heterozygosity in a thymic

N E W S A N D V I E W S

NATURE IMMUNOLOGY VOLUME 4 NUMBER 6 JUNE 2003 513

Figure 1 Bcl11a and Bcl11b in lymphoid development. Bcl11a is a transcription factor required in thedevelopment of pre-pro-B cells. Its targets could include Notch or Bcl6. Bcl11b is a close paralog butis required in αβ T cell development before the DN3 stage and full rearrangement of the TCRβ locus.Its targets could include p53, E2a, FADD, Wnt or Notch pathways. Neither Bcl11a nor Bcl11b isrequired to generate γδT cells, whereas NK cells were not examined. The derivation of thymocytes froma common lymphoid progenitor in bone marrow, as shown here, has recently been challenged, so thethymic progenitor may branch off earlier.

Common Commonlymphoid progenitor

B lcel

αβαβ TTT cellT cel

γδγδ T cell (NK?)T celll (NK?)ll (NK?)ll (N

Bcl11b

Antiapoptic, p53,AE2A, FADD, Wnt,E

Notch[[ [

DN3V DV-D-Dβ

Pre pro BPre-proo-B-pro

Bcl11a

Notch Bcl6[ [?

C. C

.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reim

mu

no

log

y

N E W S A N D V I E W S

514 VOLUME 4 NUMBER 6 JUNE 2003 NATURE IMMUNOLOGY

lymphoma12. Bcl11a and Bcl11b have a 61%identity in their peptide sequences, the zincfinger domains are 95% identical, and theycan both bind the same consensus sequenceand repress transcription6. Although thissuggests they might behave similarly, loss ofBcl11b interrupts αβ thymocyte develop-ment but has no effect on γδ T cells or Bcells. The thymic block is the opposite ofwhat would be predicted from the initialobservation that thymic lymphomas show aloss of Bcl11b—an increase in the numberof thymocytes would be expected. Thethymic block is potent, causing a nearlycomplete failure to generate double-positivecells. Rearrangement of the TCRβ locus inBcl11b–/– DN thymocytes is incomplete.Whereas they successfully rearranged fromD to J, they did not rearrange from V to D.Either this block could be an effect on acces-sibility of the Vβ genes or, because V-to-Drearrangement occurs after D-to-J, the cellscould die before the second rearrangement.Extensive apoptosis was observed in thesethymocytes, which, again, could haveblocked the V-to-D rearrangement orresulted from it. Thus, thymocytes may bedying at the pre-TCR checkpoint becausethey did not complete rearranging the βlocus and express surface pre-TCR. TheirDN3 arrest occurs at the same stage as inthymocytes that do not rearrange their βlocus, such as in Rag1–/– mice.

If thymocytes are dying at the pre-TCRcheckpoint because they do not rearrange the

β locus, development should be rescued witha rearranged TCRβ transgene or knockout ofFADD, E2A or p53, which mediate thischeckpoint. Another possibility is that thesethymocytes prematurely activate the pre-TCR checkpoint before they have time torearrange V to D. If the latter were true, theymay be rescued by knockout of FADD, E2Aor p53, but not a rearranged TCRβ trans-gene. On the other hand, if a Bcl2 transgenerescued development, it would implicate aslightly earlier apoptotic problem, for exam-ple, in IL-7 receptor signaling.

Some data5 suggested that the deathpathway in Bcl11b knockout thymocytesinvolved p53. The evidence for this was thatp53 deletion rescued embryonic thymo-cytes from massive apoptosis. But the studywas not continued to show whether p53deletion actually restored thymopoiesis.The expression of Bcl-2 and Bax, which arethe primary apoptosis regulators at thisstage, did not seem to be affected in theknockout.

The direct gene targets of Bcl11b have yetto be determined. They could be involved insignaling by Notch because its conditionalknockout in pre-T cells also resulted in ablock at V-Dβ rearrangement. The targetscould also be involved in signaling by FADD,E2A or Wnt, which are critical at this stage ofthymopoiesis. The intracellular location ofBcl11b in thymocytes is unexpected. In neu-rons, it was found in nuclei, as expected for atranscription factor, but in thymocytes,

most Bcl11a was in the mitochondria.Although the small amount in the nucleusmay well carry out all the important func-tions, could Bcl11b also have a function inmitochondria, such as protection fromapoptosis? Perhaps this resembles anotherKruppel-like ‘transcription’ factor, Aiolos,which is reported to affect apoptosis in twoopposing ways. In IL-2-dependent cells,Aiolos is protective and acts as a proper tran-scription factor, inducing Bcl-2 expressionin response to IL-2 stimulation. In IL-4-dependent cells, however, withdrawal of IL-4induces Aiolos binding to Bcl-xL, leading tocell death.

In conclusion, these knockouts of Bcl11aand Bcl11b clearly establish their require-ment at different stages of lymphopoiesis.Although this determines when they act, howthey actually work is still a mystery.

1. Kaczynski, J., Cook, T. & Urrutia, R. Genome Biol. 4,206 (2003).

2. Georgopoulos, K. Nat. Rev. Immunol. 2, 162–174(2002).

3. Buckley, A.F., Kuo, C.T. & Leiden, J.M. Nat.Immunol. 2, 698–704 (2001).

4. Liu, P. et al. Nat. Immunol. 4, 527–534 (2003).5. Wakabayashi, Y. et al. Nat. Immunol. 4, 535–541

(2003).6. Avram, D. et al. Biochem. J. 368, 555–563 (2002).7. Li, J. et al. Nat. Genet. 23, 348–353 (1999).8. Nakamura, T. et al. Mol. Cell. Biol. 20, 3178–3186

(2000).9. Avram, D. et al. J. Biol. Chem. 275, 10315–10322

(2000).10. Satterwhite, E. et al. Blood 98, 3413–3420 (2001).11. Seyfert, B.L., Allman, D., He, Y. & Staudt, L.M.

Oncogene 12, 2331–2343 (1996).12. Shinbo, T. et al. Oncogene 18, 4131–4136 (1999).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reim

mu

no

log

y